A MULTICENTER evaluation study presented at WCO 2025 has shown that artificial intelligence (AI) can significantly improve the detection of osteoporotic vertebral compression fractures (VCFs), a commonly underdiagnosed condition that contributes substantially to morbidity in aging populations.
The study compared standard clinical practice with an AI-powered screening tool, CINA-VCF Quantix, developed by Avicenna.AI. Researchers analyzed 2,224 CT scans from four major hospitals in Marseille, France, involving patients aged 50 years and older. These scans were initially performed for various clinical reasons, unrelated to suspected fractures.
Out of 489 cases flagged as positive by the AI, 312 were confirmed as true VCFs by expert radiologists, yielding a positive predictive value of 63.8%. Notably, nearly a third (31.1%) of the AI-detected fractures were entirely missed in the original radiology reports, highlighting a significant diagnostic blind spot in routine care. The AI’s false positive rate was 36.2%, primarily due to anatomical anomalies like Schmorl’s nodes or imaging artifacts.
Per-vertebra agreement between the AI and expert reviewers was impressively high at 95.8%, ranging from 80.3% for the L5 vertebra to 99.0% for L2. Moreover, the study found that screening just 14.6 CT scans with the AI tool could lead to the detection of one additional moderate-to-severe VCF that would otherwise go unreported.
While the results affirm the promise of AI in enhancing fracture detection and supporting earlier osteoporosis care, the authors emphasized the ongoing need for radiologist oversight to validate AI findings and minimize misclassification. Future efforts will explore the tool’s integration into real-time clinical workflows and its impact on treatment outcomes and healthcare efficiency.
Aleksandra Zurowska, EMJ
Reference
Ayobi A et al. Leveraging artificial intelligence for enhanced detection of osteoporotic vertebral compression fractures: a multicenter evaluation. Abstract P110. WCO Congress, 10-13 April 2025.
